How does a virtual rail behave differently from a primary rail inside a shutdown domain, using the Mychip PD_COP example?

Short Answer

Inside PD_COP the cells run on VDD1p0_SW, a virtual rail that exists only while the header switches conduct, while VDD1p0 outside stays flat at 1.0 V. When the switches open, VDD1p0_SW decays slowly as leakage drains it and internal nodes float, and even when on it sits below VDD1p0 by the switch IR drop. Isolation, retention and always-on cells in PD_COP must therefore tap VDD1p0, never the virtual rail.

Technical Explanation

• Primary rail: VDD1p0 comes from the package, powers PD_MYCHIP and feeds the switch inputs; it never turns off.
• Virtual rail: VDD1p0_SW is the switch output and PD_COP primary; it is driven only while the switches conduct.
• Turn-off: with no driver, VDD1p0_SW decays at a rate set by leakage and rail capacitance, so logic passes through undefined levels.
• Turn-on: it ramps as the chain closes and draws in-rush current from VDD1p0 that always-on neighbours can feel.
• When on: it sits I×R_on below VDD1p0, so PD_COP timing and IR must use the reduced voltage.
• Special cells: isolation, retention shadow latches and always-on buffers inside PD_COP take VDD1p0 through a secondary PG pin.
• Legacy form (still accepted by ICC2/PT): set_domain_supply_net PD_COP -primary_power_net VDD1p0_SW -primary_ground_net VSS (UPF).

# [UPF]  mychip.upf
create_supply_set SS_COP -function {power VDD1p0_SW} -function {ground VSS}
create_power_domain PD_COP -elements {U_COP} -supply {primary SS_COP}
create_power_switch COP_SW -domain PD_COP -input_supply_port {in VDD1p0} -output_supply_port {out VDD1p0_SW} -control_port {ctrl PSE} -on_state {ON in {ctrl}}
# [ICC2]  icc2_shell
report_power_domains
analyze_rail -voltage_drop static -nets {VDD1p0 VDD1p0_SW VSS}

What To Check

• PD_COP primary is VDD1p0_SW, and the switch input is VDD1p0.
• Every isolation, retention and always-on cell in PD_COP has its backup pin on VDD1p0.
• Static IR on VDD1p0_SW includes the switch drop and stays inside the budget.
• PD_COP timing uses the virtual-rail voltage, not the nominal 1.0 V.

Healthy, Suspicious & Hard-stop Results

• Healthy (illustrative): VDD1p0_SW sits at 0.973 V at peak current, a 27 mV drop inside a 30 mV budget.
• Suspicious (illustrative): VDD1p0_SW shows 45 mV of drop in one corner of VA_COP where switch density is low.
• Hard stop: A retention flop or isolation cell has its backup pin on VDD1p0_SW, so it loses its value or clamp the moment PD_COP turns off.

What The Interviewer Is Testing

• Whether you know which cells may sit on the virtual rail and which must not.
• Whether you account for switch IR and decay when you reason about a gated domain.

Follow-up Question & Model Response

"Can you power PD_COP back on before VDD1p0_SW has fully decayed?"

Candidate Model Response: Electrically yes: the switches recharge the rail from wherever it sits. The logic state is still undefined because nodes floated, so you still restore from retention and reset what was not retained. Keep isolation on until the ack arrives. The in-rush is smaller because the rail needs less charge.